6 files changed, 215 insertions, 34 deletions

diff --git a/src/core/core.cpp b/src/core/core.cpp
index 5d4ecdce5..fd1bdcaf0 100644
--- a/src/core/core.cpp
+++ b/src/core/core.cpp
@@ -149,6 +149,9 @@ struct System::Impl {
 
         device_memory = std::make_unique<Core::DeviceMemory>(system);
 
+        kernel.SetMulticore(Settings::values.use_multi_core);
+        cpu_manager.SetMulticore(Settings::values.use_multi_core);
+
         core_timing.Initialize([&system]() { system.RegisterHostThread(); });
         kernel.Initialize();
         cpu_manager.Initialize();
diff --git a/src/core/cpu_manager.cpp b/src/core/cpu_manager.cpp
index 9a261968a..e72f89808 100644
--- a/src/core/cpu_manager.cpp
+++ b/src/core/cpu_manager.cpp
@@ -26,9 +26,13 @@ void CpuManager::ThreadStart(CpuManager& cpu_manager, std::size_t core) {
 
 void CpuManager::Initialize() {
     running_mode = true;
-    for (std::size_t core = 0; core < Core::Hardware::NUM_CPU_CORES; core++) {
-        core_data[core].host_thread =
-            std::make_unique<std::thread>(ThreadStart, std::ref(*this), core);
+    if (is_multicore) {
+        for (std::size_t core = 0; core < Core::Hardware::NUM_CPU_CORES; core++) {
+            core_data[core].host_thread =
+                std::make_unique<std::thread>(ThreadStart, std::ref(*this), core);
+        }
+    } else {
+        core_data[0].host_thread = std::make_unique<std::thread>(ThreadStart, std::ref(*this), 0);
     }
 }
 
@@ -41,52 +45,72 @@ void CpuManager::Shutdown() {
     }
 }
 
+std::function<void(void*)> CpuManager::GetGuestThreadStartFunc() {
+    return std::function<void(void*)>(GuestThreadFunction);
+}
+
+std::function<void(void*)> CpuManager::GetIdleThreadStartFunc() {
+    return std::function<void(void*)>(IdleThreadFunction);
+}
+
+std::function<void(void*)> CpuManager::GetSuspendThreadStartFunc() {
+    return std::function<void(void*)>(SuspendThreadFunction);
+}
+
 void CpuManager::GuestThreadFunction(void* cpu_manager_) {
     CpuManager* cpu_manager = static_cast<CpuManager*>(cpu_manager_);
-    cpu_manager->RunGuestThread();
+    if (cpu_manager->is_multicore) {
+        cpu_manager->MultiCoreRunGuestThread();
+    } else {
+        cpu_manager->SingleCoreRunGuestThread();
+    }
 }
 
 void CpuManager::GuestRewindFunction(void* cpu_manager_) {
     CpuManager* cpu_manager = static_cast<CpuManager*>(cpu_manager_);
-    cpu_manager->RunGuestLoop();
+    if (cpu_manager->is_multicore) {
+        cpu_manager->MultiCoreRunGuestLoop();
+    } else {
+        cpu_manager->SingleCoreRunGuestLoop();
+    }
 }
 
 void CpuManager::IdleThreadFunction(void* cpu_manager_) {
     CpuManager* cpu_manager = static_cast<CpuManager*>(cpu_manager_);
-    cpu_manager->RunIdleThread();
+    if (cpu_manager->is_multicore) {
+        cpu_manager->MultiCoreRunIdleThread();
+    } else {
+        cpu_manager->SingleCoreRunIdleThread();
+    }
 }
 
 void CpuManager::SuspendThreadFunction(void* cpu_manager_) {
     CpuManager* cpu_manager = static_cast<CpuManager*>(cpu_manager_);
-    cpu_manager->RunSuspendThread();
-}
-
-std::function<void(void*)> CpuManager::GetGuestThreadStartFunc() {
-    return std::function<void(void*)>(GuestThreadFunction);
-}
-
-std::function<void(void*)> CpuManager::GetIdleThreadStartFunc() {
-    return std::function<void(void*)>(IdleThreadFunction);
-}
-
-std::function<void(void*)> CpuManager::GetSuspendThreadStartFunc() {
-    return std::function<void(void*)>(SuspendThreadFunction);
+    if (cpu_manager->is_multicore) {
+        cpu_manager->MultiCoreRunSuspendThread();
+    } else {
+        cpu_manager->SingleCoreRunSuspendThread();
+    }
 }
 
 void* CpuManager::GetStartFuncParamater() {
     return static_cast<void*>(this);
 }
 
-void CpuManager::RunGuestThread() {
+///////////////////////////////////////////////////////////////////////////////
+///                             MultiCore                                   ///
+///////////////////////////////////////////////////////////////////////////////
+
+void CpuManager::MultiCoreRunGuestThread() {
     auto& kernel = system.Kernel();
     {
         auto& sched = kernel.CurrentScheduler();
         sched.OnThreadStart();
     }
-    RunGuestLoop();
+    MultiCoreRunGuestLoop();
 }
 
-void CpuManager::RunGuestLoop() {
+void CpuManager::MultiCoreRunGuestLoop() {
     auto& kernel = system.Kernel();
     auto* thread = kernel.CurrentScheduler().GetCurrentThread();
     auto host_context = thread->GetHostContext();
@@ -103,7 +127,7 @@ void CpuManager::RunGuestLoop() {
     }
 }
 
-void CpuManager::RunIdleThread() {
+void CpuManager::MultiCoreRunIdleThread() {
     auto& kernel = system.Kernel();
     while (true) {
         auto& physical_core = kernel.CurrentPhysicalCore();
@@ -113,7 +137,7 @@ void CpuManager::RunIdleThread() {
     }
 }
 
-void CpuManager::RunSuspendThread() {
+void CpuManager::MultiCoreRunSuspendThread() {
     auto& kernel = system.Kernel();
     {
         auto& sched = kernel.CurrentScheduler();
@@ -130,7 +154,7 @@ void CpuManager::RunSuspendThread() {
     }
 }
 
-void CpuManager::Pause(bool paused) {
+void CpuManager::MultiCorePause(bool paused) {
     if (!paused) {
         bool all_not_barrier = false;
         while (!all_not_barrier) {
@@ -171,10 +195,120 @@ void CpuManager::Pause(bool paused) {
     paused_state = paused;
 }
 
+///////////////////////////////////////////////////////////////////////////////
+///                             SingleCore                                   ///
+///////////////////////////////////////////////////////////////////////////////
+
+void CpuManager::SingleCoreRunGuestThread() {
+    auto& kernel = system.Kernel();
+    {
+        auto& sched = kernel.CurrentScheduler();
+        sched.OnThreadStart();
+    }
+    SingleCoreRunGuestLoop();
+}
+
+void CpuManager::SingleCoreRunGuestLoop() {
+    auto& kernel = system.Kernel();
+    auto* thread = kernel.CurrentScheduler().GetCurrentThread();
+    auto host_context = thread->GetHostContext();
+    host_context->SetRewindPoint(std::function<void(void*)>(GuestRewindFunction), this);
+    host_context.reset();
+    while (true) {
+        auto& physical_core = kernel.CurrentPhysicalCore();
+        while (!physical_core.IsInterrupted()) {
+            physical_core.Run();
+            preemption_count++;
+            if (preemption_count % max_cycle_runs == 0) {
+                break;
+            }
+        }
+        physical_core.ClearExclusive();
+        PreemptSingleCore();
+        auto& scheduler = physical_core.Scheduler();
+        scheduler.TryDoContextSwitch();
+    }
+}
+
+void CpuManager::SingleCoreRunIdleThread() {
+    auto& kernel = system.Kernel();
+    while (true) {
+        auto& physical_core = kernel.CurrentPhysicalCore();
+        PreemptSingleCore();
+        auto& scheduler = physical_core.Scheduler();
+        scheduler.TryDoContextSwitch();
+    }
+}
+
+void CpuManager::SingleCoreRunSuspendThread() {
+    auto& kernel = system.Kernel();
+    {
+        auto& sched = kernel.CurrentScheduler();
+        sched.OnThreadStart();
+    }
+    while (true) {
+        auto core = kernel.GetCurrentHostThreadID();
+        auto& scheduler = kernel.CurrentScheduler();
+        Kernel::Thread* current_thread = scheduler.GetCurrentThread();
+        Common::Fiber::YieldTo(current_thread->GetHostContext(), core_data[0].host_context);
+        ASSERT(scheduler.ContextSwitchPending());
+        ASSERT(core == kernel.GetCurrentHostThreadID());
+        scheduler.TryDoContextSwitch();
+    }
+}
+
+void CpuManager::PreemptSingleCore() {
+    preemption_count = 0;
+    std::size_t old_core = current_core;
+    current_core = (current_core + 1) % Core::Hardware::NUM_CPU_CORES;
+    auto& scheduler = system.Kernel().Scheduler(old_core);
+    Kernel::Thread* current_thread = system.Kernel().Scheduler(old_core).GetCurrentThread();
+    Kernel::Thread* next_thread = system.Kernel().Scheduler(current_core).GetCurrentThread();
+    Common::Fiber::YieldTo(current_thread->GetHostContext(), next_thread->GetHostContext());
+}
+
+void CpuManager::SingleCorePause(bool paused) {
+    if (!paused) {
+        bool all_not_barrier = false;
+        while (!all_not_barrier) {
+            all_not_barrier = !core_data[0].is_running.load() && core_data[0].initialized.load();
+        }
+        core_data[0].enter_barrier->Set();
+        if (paused_state.load()) {
+            bool all_barrier = false;
+            while (!all_barrier) {
+                all_barrier = core_data[0].is_paused.load() && core_data[0].initialized.load();
+            }
+            core_data[0].exit_barrier->Set();
+        }
+    } else {
+        /// Wait until all cores are paused.
+        bool all_barrier = false;
+        while (!all_barrier) {
+            all_barrier = core_data[0].is_paused.load() && core_data[0].initialized.load();
+        }
+        /// Don't release the barrier
+    }
+    paused_state = paused;
+}
+
+void CpuManager::Pause(bool paused) {
+    if (is_multicore) {
+        MultiCorePause(paused);
+    } else {
+        SingleCorePause(paused);
+    }
+}
+
 void CpuManager::RunThread(std::size_t core) {
     /// Initialization
     system.RegisterCoreThread(core);
-    std::string name = "yuzu:CoreHostThread_" + std::to_string(core);
+    std::string name;
+    if (is_multicore) {
+        name = "yuzu:CoreCPUThread_" + std::to_string(core);
+    } else {
+        name = "yuzu:CPUThread";
+    }
     MicroProfileOnThreadCreate(name.c_str());
     Common::SetCurrentThreadName(name.c_str());
     auto& data = core_data[core];
diff --git a/src/core/cpu_manager.h b/src/core/cpu_manager.h
index e83ab20f9..1e81481ec 100644
--- a/src/core/cpu_manager.h
+++ b/src/core/cpu_manager.h
@@ -30,6 +30,10 @@ public:
     CpuManager& operator=(const CpuManager&) = delete;
     CpuManager& operator=(CpuManager&&) = delete;
 
+    /// Sets if emulation is multicore or single core, must be set before Initialize
+    void SetMulticore(bool is_multicore) {
+        this->is_multicore = is_multicore;
+    }
     void Initialize();
     void Shutdown();
 
@@ -40,21 +44,34 @@ public:
     std::function<void(void*)> GetSuspendThreadStartFunc();
     void* GetStartFuncParamater();
 
+    std::size_t CurrentCore() const {
+        return current_core;
+    }
+
 private:
     static void GuestThreadFunction(void* cpu_manager);
     static void GuestRewindFunction(void* cpu_manager);
     static void IdleThreadFunction(void* cpu_manager);
     static void SuspendThreadFunction(void* cpu_manager);
 
-    void RunGuestThread();
-    void RunGuestLoop();
-    void RunIdleThread();
-    void RunSuspendThread();
+    void MultiCoreRunGuestThread();
+    void MultiCoreRunGuestLoop();
+    void MultiCoreRunIdleThread();
+    void MultiCoreRunSuspendThread();
+    void MultiCorePause(bool paused);
+
+    void SingleCoreRunGuestThread();
+    void SingleCoreRunGuestLoop();
+    void SingleCoreRunIdleThread();
+    void SingleCoreRunSuspendThread();
+    void SingleCorePause(bool paused);
 
     static void ThreadStart(CpuManager& cpu_manager, std::size_t core);
 
     void RunThread(std::size_t core);
 
+    void PreemptSingleCore();
+
     struct CoreData {
         std::shared_ptr<Common::Fiber> host_context;
         std::unique_ptr<Common::Event> enter_barrier;
@@ -70,6 +87,11 @@ private:
 
     std::array<CoreData, Core::Hardware::NUM_CPU_CORES> core_data{};
 
+    bool is_multicore{};
+    std::size_t current_core{};
+    std::size_t preemption_count{};
+    static constexpr std::size_t max_cycle_runs = 5;
+
     System& system;
 };
 
diff --git a/src/core/hle/kernel/kernel.cpp b/src/core/hle/kernel/kernel.cpp
index 721ab1e70..4a091ea38 100644
--- a/src/core/hle/kernel/kernel.cpp
+++ b/src/core/hle/kernel/kernel.cpp
@@ -113,6 +113,10 @@ struct KernelCore::Impl {
     explicit Impl(Core::System& system, KernelCore& kernel)
         : global_scheduler{kernel}, synchronization{system}, time_manager{system}, system{system} {}
 
+    void SetMulticore(bool is_multicore) {
+        this->is_multicore = is_multicore;
+    }
+
     void Initialize(KernelCore& kernel) {
         Shutdown();
 
@@ -237,6 +241,9 @@ struct KernelCore::Impl {
 
     void RegisterCoreThread(std::size_t core_id) {
         std::unique_lock lock{register_thread_mutex};
+        if (!is_multicore) {
+            single_core_thread_id = std::this_thread::get_id();
+        }
         const std::thread::id this_id = std::this_thread::get_id();
         const auto it = host_thread_ids.find(this_id);
         ASSERT(core_id < Core::Hardware::NUM_CPU_CORES);
@@ -258,6 +265,11 @@ struct KernelCore::Impl {
 
     u32 GetCurrentHostThreadID() const {
         const std::thread::id this_id = std::this_thread::get_id();
+        if (!is_multicore) {
+            if (single_core_thread_id == this_id) {
+                return static_cast<u32>(system.GetCpuManager().CurrentCore());
+            }
+        }
         const auto it = host_thread_ids.find(this_id);
         if (it == host_thread_ids.end()) {
             return Core::INVALID_HOST_THREAD_ID;
@@ -378,6 +390,9 @@ struct KernelCore::Impl {
 
     std::array<std::shared_ptr<Thread>, Core::Hardware::NUM_CPU_CORES> suspend_threads{};
 
+    bool is_multicore{};
+    std::thread::id single_core_thread_id{};
+
     // System context
     Core::System& system;
 };
@@ -387,6 +402,10 @@ KernelCore::~KernelCore() {
     Shutdown();
 }
 
+void KernelCore::SetMulticore(bool is_multicore) {
+    impl->SetMulticore(is_multicore);
+}
+
 void KernelCore::Initialize() {
     impl->Initialize(*this);
 }
diff --git a/src/core/hle/kernel/kernel.h b/src/core/hle/kernel/kernel.h
index 5d32a8329..162bbd2f8 100644
--- a/src/core/hle/kernel/kernel.h
+++ b/src/core/hle/kernel/kernel.h
@@ -65,6 +65,9 @@ public:
     KernelCore(KernelCore&&) = delete;
     KernelCore& operator=(KernelCore&&) = delete;
 
+    /// Sets if emulation is multicore or single core, must be set before Initialize
+    void SetMulticore(bool is_multicore);
+
     /// Resets the kernel to a clean slate for use.
     void Initialize();
 
diff --git a/src/core/memory.cpp b/src/core/memory.cpp
index 4cb5d05e5..7def00768 100644
--- a/src/core/memory.cpp
+++ b/src/core/memory.cpp
@@ -715,8 +715,8 @@ struct Memory::Impl {
             ASSERT_MSG(false, "Mapped memory page without a pointer @ {:016X}", vaddr);
             break;
         case Common::PageType::RasterizerCachedMemory: {
-            u8* host_ptr{GetPointerFromVMA(vaddr)};
-            system.GPU().InvalidateRegion(ToCacheAddr(host_ptr), sizeof(T));
+            u8* host_ptr{GetPointerFromRasterizerCachedMemory(vaddr)};
+            system.GPU().InvalidateRegion(vaddr, sizeof(T));
             T volatile* pointer = reinterpret_cast<T volatile*>(&host_ptr);
             return Common::AtomicCompareAndSwap(pointer, data, expected);
             break;
@@ -745,8 +745,8 @@ struct Memory::Impl {
             ASSERT_MSG(false, "Mapped memory page without a pointer @ {:016X}", vaddr);
             break;
         case Common::PageType::RasterizerCachedMemory: {
-            u8* host_ptr{GetPointerFromVMA(vaddr)};
-            system.GPU().InvalidateRegion(ToCacheAddr(host_ptr), sizeof(u128));
+            u8* host_ptr{GetPointerFromRasterizerCachedMemory(vaddr)};
+            system.GPU().InvalidateRegion(vaddr, sizeof(u128));
             u64 volatile* pointer = reinterpret_cast<u64 volatile*>(&host_ptr);
             return Common::AtomicCompareAndSwap(pointer, data, expected);
             break;